OpClave sends your question to all four models in parallel, then a synthesis model called Voyance reads all four responses and produces one answer. The system is set up so that it notes where they agree, where they disagree, and why.
A few things that make it different:
- Models are anonymized before synthesis so Voyance doesn't bias toward any one of them - File attachments. upload a PDF or image and all four models reason over it using an extract-once architecture so it doesn't cost 4x - You can toggle models on/off and pick which model leads the synthesis
One use case I've found personally useful: using it to generate prompts before feeding them into tools like Lovable or Claude. Four models debating how to approach a problem produces significantly better prompts than writing one yourself or by just simply asking one AI to write one for you.
Free to use, no account needed.
opclave.com
Happy to answer questions about the architecture
built with FastAPI, SSE streaming, and the Anthropic/OpenAI/Google/xAI APIs.